Machine learning applied to X-ray tomography as a new tool to analyze the voids in RRP Nb₃Sn wires

Area of Science:

• Materials Science
• Particle Physics
• Superconductivity

Background:

• High-performance Restacked-Rod-Process (RRP) Nb3Sn wires are crucial for developing compact magnets exceeding 15 Tesla for future particle physics experiments.
• Understanding the electro-mechanical and electro-thermal properties of these wires is essential for their successful application.

Purpose of the Study:

• To introduce a novel analytical tool combining X-ray micro-tomography and unsupervised machine learning for characterizing internal features of RRP wires.
• To identify and analyze void distribution and morphology within RRP wires, which impact their performance.

Main Methods:

• Utilizing X-ray micro-tomography to image the internal structure of RRP Nb3Sn wires.
• Applying an unsupervised machine learning algorithm to analyze the tomographic data and identify void characteristics.
• Categorizing voids into two types: those within the copper matrix and those within the Nb3Sn sub-elements.

Main Results:

• Successfully detected and characterized two distinct types of voids in RRP Nb3Sn wires.
• Identified voids in the copper matrix linked to tin (Sn) leakage, negatively affecting electro-thermal stability.
• Identified voids within Nb3Sn sub-elements, detrimental to electro-mechanical performance under high stress conditions.

Conclusions:

• The developed X-ray tomography analysis tool offers a powerful method for studying RRP wire internal features.
• This tool can aid in modeling and predicting the electro-mechanical and electro-thermal behavior of RRP wires.
• Optimization of RRP wire design for enhanced performance in high-field applications is facilitated by this advanced characterization technique.